 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
Car Forum / Driving, Maintenance, Tuning / Car Audio / March 2004Speaker feedback/feedforward control
I've got a question here that I think I'm right about, but need some verification. Here's the deal: I have a lab class in which one of the labs takes a speaker (DVC 6") and tries to extended the lower cutoff frequency by half an octave. Here's what we did: We connected one voice coil to an amp and an oscilloscope, and connected the other voice coil to the oscilloscope in order to measure the input and the output (cone motion measure in volts). We got a good representation of the frequency response (which is a really pathetic response). We then derived a transfer funciton to accurately model the frequency response. With this, we could utilize feedforward or feedback/feedforward control to extend the cutoff frequency. We decided to use a simple differential amplifier from an op-amp which falls into the feedback/feedforward category by how we used it. With all the stupid things about this lab aside (trust me, there are alot), here's my question. Wouldn't using any type of feedback from the speaker completely screw up the output due to the fact that music is transient and the feedback would be correcting amplitude deficiencies that have already passed? Simpler terms...the motion of the cone is measured, inputted back into the circuit, and this information is, in effect, used to 'amplify' some frequencies to extend the cutoff. Am I right? Thanks Gatti > I've got a question here that I think I'm right about, but need some > verification. Here's the deal: I have a lab class in which one of the [quoted text clipped - 18 lines] > information is, in effect, used to 'amplify' some frequencies to > extend the cutoff. Am I right? Thanks Ah, a question about the causal nature of feedback. :) Ok, so I'm not entirely clear on your implementation of the feedback circuit. You're using the second coil to measure the cone motion (I'm assuming you're using the proper transform in order to arrive at a value for cone displacement rather than cone velocity). Then you're plugging that into a circuit to modify some sort of filter properties of the input signal, no? Anyway, your suspicions are probably correct. Keep in mind that the speaker coil has a large reactance, so you're right that your feedback will significantly lag the output signal. Honestly, I don't know how to go about implementing the servo control you're suggesting. Anyone familiar with Velodyne's strategy may be able to provide more insight than I can. > You're using the second coil to measure the cone motion (I'm assuming you're using the proper transform in order to arrive at a value for cone displacement rather than cone velocity). We normalize the output by the input and got a magnitude (-dB due to losses) for each measured frequency. > Then you're plugging that into a circuit to modify some sort of filter properties of the input signal, no? We made a differential amplifer that has one input of our original signal and the other input of the output (measured from the passive voice coil). The ouput of the differential amp is sent to the active voice coil. > Keep in mind that the speaker coil has a large reactance, so you're right that your feedback will significantly lag the output signal. That's exactly what I was thinking. > Honestly, I don't know how to go about implementing the servo control you're suggesting. Again, exactly what I was thinking. Not sure about this...but isn't feedback (differential amp in our case) used for modifying time variant signals and not frequency variant signals? I guess my question was getting at the fact that this application is completely wrong for feedback/feedforward control. My professor made up the lab, and I'd really like to rip on him about this, so I needed my thinking verified. Thanks. Gatti > > You're using the second coil to measure the cone motion (I'm assuming you're using the proper transform in order to arrive at a value for cone displacement rather than cone velocity). > We normalize the output by the input and got a magnitude (-dB due to > losses) for each measured frequency. [quoted text clipped - 5 lines] > voice coil). The ouput of the differential amp is sent to the active > voice coil. Yikes. You sure you explained that right? Or am I misunderstanding you? You have to realize what you're measuring on the passive coil. You're measuring cone velocity. This is not the same as measuring cone output. Such a setup would naturally have an enhanced output for higher frequencies (cone moves faster). > > Keep in mind that the speaker coil has a large reactance, so you're right that your feedback will significantly lag the output signal. > [quoted text clipped - 9 lines] > really like to rip on him about this, so I needed my thinking > verified. Thanks. Well, many college professors are notorious for being both cocky and uninformed, so take his advice for what it's worth. :) However, I feel like I'm still in the dark about your project. First of all, it appears you're using velocity rather than cone motion as your control. You basically need to add an integrator to turn dx/dt into x. Secondly, as you've acknowledged, your feedback circuit is limited by the delay. But it'll still work if you're giving it a constant signal. > Yikes. You sure you explained that right? Or am I misunderstanding you? > You have to realize what you're measuring on the passive coil. You're > measuring one velocity. This is not the same as measuring cone output. > Such a setup would naturally have an enhanced output for higher frequencies > (cone moves faster). Okay...I'm a little foggy on what we measured. We connected a signal generator to an amp (as well as one input of the oscope), then to one voice coil. We connected the other voice coil to another input on the oscope. Based on the shape of the response we got, I'm assuming we measured cone output. There is a peak around 70Hz (resonance frequency), then a sever drop at about 200Hz (impedance spike I guess, its a Radio Shack speaker), and then comes back up and levels off around 1000Hz. The resonance frequency has nearly the same magnitude as the higher frequencies where it levels off. So, based on that, I thought we were measuring cone output, is this correct? How would we measure cone output if this isn't the case? > > Yikes. You sure you explained that right? Or am I misunderstanding you? > > You have to realize what you're measuring on the passive coil. You're [quoted text clipped - 13 lines] > thought we were measuring cone output, is this correct? How would we > measure cone output if this isn't the case? What you measure from the second coil is the voltage induced by moving a coil in a constant magnetic field. The voltage isn't proportional to the excursion of the coil (and cone, of course, since they're connected). It's proportional to the speed of the coil. So you're measuring cone speed rather than cone excursion - in other words, you're measuring the time derivative of cone excursion when you measure cone velocity. I suspect you want cone excursion, so you would need to account for that by integrating. Turns out the frequency is the same for both (the derivative of a sine is a cosine), but the amplitude will be different. So if you play a 200 Hz signal with the same amplitude as a 2kHz signal, the 2kHz signal will generally produce a higher voltage due to higher velocity. This is further complicated by the fact that you're not really inducing a voltage by moving the coil. Rather, you're inducing a current. So the impedance of the coil must also be taken into account. In the end, the coil is probably serving to attenuate high frequencies while the cone velocity that you're measuring is enhancing high frequencies. So the effect you're seeing is being counteracted at high frequencies, but not entirely. That's why it appears to be working for you. But, again, if cone motion is what you're after exactly, you'll have to introduce integration into your circuit and compensation for the coil impedance. Im no expert but I think theres more to it than just receiving the feedback signal and re-amplifying that into the mix... Part of your circuit needs to analyze and compare the original signal and the feedback signal and mix the two optimally. I think this is the basic idea.... And yes transient response is also a key factor. Sounds like a cool lab what class is it for? Garrett > I've got a question here that I think I'm right about, but need some > verification. Here's the deal: I have a lab class in which one of the [quoted text clipped - 20 lines] > > Gatti > Im no expert but I think theres more to it than just receiving the feedback > signal and re-amplifying that into the mix... Part of your circuit needs to > analyze and compare the original signal and the feedback signal and mix the > two optimally. I think this is the basic idea.... And yes transient > response is also a key factor. Yeah, that's what our differential amplifier is for. This is separate from the amplifier of the signal. The differential amp has one input of the original signal and the other input of the feedback signal, and performs its operation on these two signals. > Sounds like a cool lab what class is it for? Mechanical Engineering 495, University of Michigan. It's a lab class. We are given a scenario letter and go from there with some direction from professors. This setup was given to us and we were told to use it. But, if what I'm thinking is correct, there would be no use for this type of feedback for transient signals (i.e. all music). > the speaker completely screw up the output due to the fact that music > is transient and the feedback would be correcting amplitude > deficiencies that have already passed? Simpler terms...the motion of > the cone is measured, inputted back into the circuit In practice it works at subwoofer frequencies and a few manufacturer actually build them, namely Velodyne and Paradigm. They even go as far as using a physical accelerometer (SP?) to got their correcting signal. But I haven't seen such design yet used for full range speakers.  SignatureEric (Dero) Desrochers Hiroshima 45, Tchernobyl 86, Windows 95 > In practice it works at subwoofer frequencies and a few manufacturer > actually build them, namely Velodyne and Paradigm. They even go as far > as using a physical accelerometer (SP?) to got their correcting signal. > > But I haven't seen such design yet used for full range speakers. Isn't this the same idea as Philips MFB (motional feedback)? And it is used for correcting distortion which is time dependent (effectively constant frequency?), so I understand why this works. But for extending the bandwidth of a speaker, which is frequency dependent (and constantly changing for music), I can't see feedback working. I'm sorry if I've beaten this to death, just trying to get things straight. Also, wouldn't only having one voice coil affect the frequency response of the speaker as well as the T/S parameters? So, I was thinking we could probably get the extended bandwidth my professor is after (and a better frequency response) if we just used the two voice coils like they were intended. Is this correct? Thanks again guys. Hey! I'm not a specialist of those systems! I do know that motion sensing and correction do works at subwoofer frequencies, ie < 100 Hz or so, as several commercially successsful examples exists in the home audio market. I also know that distorsion correction by the mean of negative feedback is used in almost every audio amplifiers for as long as audio amplifiers existed. Actually, negative feedback provide an amplifier circuit with those outstanding enhancement : Considerable bandwith augmentation (like going from 15 Hz to 40 kHz) Distorsion reduction Noise reduction Damping factor augmentation Better stability Etc At the cost of reducing the global gain. Since several amplifiers have bandwith in the several hundred kHz range, it's clear that the feedback circuit actually DO work, even at very high frequencies. You should not forget that the electrons in an electrical circuit are VERY fast so reacting to a continously variable signal below 20 kHz is trivial. SignatureEric (Dero) Desrochers Hiroshima 45, Tchernobyl 86, Windows 95 > At the cost of reducing the global gain. > [quoted text clipped - 3 lines] > electrical circuit are VERY fast so reacting to a continously variable > signal below 20 kHz is trivial. But it's not a matter of the speed of electrons. It's an issue with the time lag due to the inductance of the voice coil. > > At the cost of reducing the global gain. > > [quoted text clipped - 6 lines] > But it's not a matter of the speed of electrons. It's an issue with the > time lag due to the inductance of the voice coil. That's correct. I somewhat thought the precedent poster said that feedback could never work with a musical signal, only with continuous tones, so I pointed out that audio amps successfully use feedback with full range musical signals. Servo systems that senses the position/velocity of a loudspeaker cone won't work at higher frequencies (for the reason you stated) but will work at low frequencies even in the case of a variable, random musical signal. SignatureEric (Dero) Desrochers Hiroshima 45, Tchernobyl 86, Windows 95 > > But it's not a matter of the speed of electrons. It's an issue with the > > time lag due to the inductance of the voice coil. [quoted text clipped - 8 lines] > work at low frequencies even in the case of a variable, random musical > signal. At least not using the second voice coil to make the measurements. I would imagine a low impedance coil would improve the timing capabilities, although the signal strength would be weaker. |
Reply to this Message |
 Sign In
 Join
 My Latest Posts My Monitored Threads My Blog My Photo Gallery My Profile My Homepage Start New Thread Enable EMail Alerts Rate this Thread
|
©2008 Advenet LLC Privacy Policy - Terms of Use
This website includes both content owned or controlled by Advenet as well as content owned or controlled by third parties.